Target object defect identification method, device and system

Abstract

The present application relates to the technical field of target object detection, and specifically relates to a target object defect identification method, device and system, and a computer-readable storage medium. In the target object defect identification method, initial image information of a target object and defect location information obtained by a defect identification model performing preliminary identification on image information are processed separately, a preliminary identification result is displayed, a user is allowed to calibrate and modify a defect identification result obtained after the preliminary identification, so as to obtain a calibrated and modified second marked image, and the second marked image is used as a final defect identification result. This approach avoids the problem in the prior art that a user is unable to independently modify a target object defect identification result when the identification result of a defect identification model is inaccurate, thereby significantly improving the accuracy and convenience of defect identification results.

Description

Target object defect identification methods, equipment and systems

[0001] This application claims priority to Chinese Patent Application No. 2024119105653, filed on December 24, 2024, entitled “Method, apparatus and system for identifying defects in target objects”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of target object detection technology, specifically to a method, device, system, and computer-readable storage medium for identifying defects in target objects. Background Technology

[0003] As people's economic living standards improve, the use of vehicles and smart devices has become more and more frequent. However, during use, collisions and scrapes are inevitable. When repairing vehicles and smart devices, or when assessing damages for insurance, it is often necessary to locate and evaluate the collisions and scrapes.

[0004] The inventors of this application discovered in their research that current car and smart device repair shops rely on paper forms to roughly inspect the appearance of cars and smart devices during repairs or insurance claims. They judge the damage such as scratches and scrapes by human eyes, and then record the inspection results on paper and manually enter them into the computer. The process is cumbersome and inefficient. Summary of the Invention

[0005] In view of the above problems, embodiments of this application provide a method, device, system and computer-readable storage medium for identifying defects in a target object, in order to solve the above-mentioned technical problems existing in the prior art.

[0006] One aspect of this application proposes a method for identifying defects in a target object, the method comprising:

[0007] Acquire image information of the target object;

[0008] The image information is sent to a pre-trained defect recognition model so that the defect recognition model can identify defects in the image information;

[0009] Receive the defect identification result generated by the defect identification model, wherein the defect identification result includes defect location information;

[0010] Based on the defect location information, the defects in the image information are identified to generate a first identified image;

[0011] In response to an operation on the first identified image, obtain the operation trajectory information corresponding to the operation;

[0012] The first identification image is identified based on the operation trajectory information to generate a second identification image;

[0013] The defects of the target object are displayed based on the second identification image.

[0014] Preferably, in some embodiments, the step of identifying the first identification image based on the operation trajectory information to generate a second identification image includes:

[0015] Multiple coordinate points corresponding to the operation trajectory information are obtained according to a preset pixel spacing threshold;

[0016] The first identification image is identified based on multiple coordinate points to generate a second identification image.

[0017] Preferably, in some embodiments, after identifying the defects in the image information based on the defect location information and generating a first identified image, the process includes:

[0018] In response to the inverted selection operation on the first identifier image, the identifier on the first identifier image is deleted.

[0019] Preferably, in some embodiments, the defect identification result further includes defect type information, which is used to indicate the defect type of the defect area corresponding to the defect location information;

[0020] The step of obtaining operation trajectory information corresponding to the operation in response to an operation on the first identifier image further includes:

[0021] Obtain the defect type information corresponding to the operation trajectory information;

[0022] The method further includes:

[0023] In response to a query operation targeting the defect type information, the defect identification results and the operation trajectory information are traversed to generate query results.

[0024] Preferably, in some embodiments, the method further includes:

[0025] The second identifier image is sent to the defect recognition model so that the defect recognition model can perform self-calibration training based on the second identifier image.

[0026] Another aspect of this application provides a target object defect identification system, the system including a target object defect identification device and a server;

[0027] The target defect identification device acquires image information of the target object and sends the image information to the server;

[0028] The server runs a pre-trained defect recognition model, which is used to identify defects in the image information and send the defect recognition results to the target object defect recognition device. The defect recognition results include defect location information.

[0029] The target object defect identification device receives the defect location information and identifies the defects in the image information according to the defect location information, generating a first identification image;

[0030] The target object defect identification device is further configured to, in response to an operation on the first identification image, acquire operation trajectory information corresponding to the operation; identify the first identification image according to the operation trajectory information to generate a second identification image; and display the target object defect according to the second identification image.

[0031] Preferably, in some embodiments, the server includes an AI server and a file server;

[0032] The target defect identification device is communicatively connected to the AI ​​server and the file server, respectively.

[0033] The target defect identification device is also used to send a pre-signed address acquisition message to the AI ​​server;

[0034] The AI ​​server sends pre-signature address information to the target defect identification device based on the pre-signature address, and the pre-signature address information includes the file server address.

[0035] The target defect identification device sends the image information to the file server according to the file server address.

[0036] Preferably, in some embodiments, the target defect identification device is further configured to send an AI analysis request to the AI ​​server in response to the AI ​​analysis operation;

[0037] The AI ​​server obtains the image information from the file server according to the AI ​​parsing request in order to identify defects in the image information.

[0038] Preferably, in some embodiments, the target defect identification device is further configured to locally store the second identification image and send the second identification image to the AI ​​server so that the defect identification model can perform self-calibration training based on the second identification image.

[0039] A third aspect of this application provides a target defect identification device, comprising: a camera, a display screen, a processor, a memory, a communication interface, and a communication bus, wherein the camera, display screen, processor, memory, and communication interface communicate with each other through the communication bus;

[0040] The camera is used to capture images of the target object and obtain its image information;

[0041] The display screen is used to provide an interactive interface with the outside world and to display the defect identification results;

[0042] The memory is used to store at least one program that causes the processor to perform the operation of the target defect identification method proposed in the above embodiments.

[0043] In a fourth aspect of this application, a computer-readable storage medium is provided, wherein at least one program is stored in the storage medium, and when the program is run on a target defect identification device, the target defect identification device performs the operation of the target defect identification method proposed in the above embodiments.

[0044] The target object defect identification method, device, and system proposed in this application, after the defect identification model completes the initial identification of image information, returns the defect location information to the target object defect identification device. The target object defect identification device then identifies the defect based on the original image information according to this defect location information. Simultaneously, the target object defect identification device allows the user to modify and supplement the identification results of the defect identification model. The target object defect identification device provides a modification interface to the user, receives the trajectory information of the user's secondary identification, and performs secondary identification based on the trajectory information on the basis of the first defect identification, forming the final target object defect identification result. This application embodiment, by improving the returned results of the defect identification model and recording the initial image information and modification identifiers separately, facilitates the modification of the defect identification results, avoiding the problem in the prior art where users cannot independently modify the target object defect identification results when the defect identification model's identification results are inaccurate, greatly improving the accuracy and convenience of defect identification results.

[0045] The above description is merely an overview of the technical solutions of the embodiments of this application. In order to better understand the technical means of the embodiments of this application and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of this application more obvious and understandable, specific implementation methods of this application are described below. Attached Figure Description

[0046] Figure 1 is a schematic diagram of the structure of a target object defect identification system proposed in an embodiment of this application;

[0047] Figure 2 is a schematic diagram of the structure of a target object defect identification device proposed in an embodiment of this application;

[0048] Figure 3 is a schematic flowchart of a target object defect identification method proposed in an embodiment of this application;

[0049] Figure 4 is a signaling diagram of a target defect identification system proposed in an embodiment of this application;

[0050] Figure 5 is a schematic diagram of the shooting interface of the target object defect identification device proposed in the embodiment of this application;

[0051] Figure 6 is a schematic diagram of the defect modification interface of the target object defect identification device proposed in the embodiment of this application;

[0052] Figure 7 is a schematic diagram of the recognition result interface of the target object defect recognition device proposed in the embodiment of this application. Detailed Implementation

[0053] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein.

[0054] With the development of artificial intelligence (AI) technology, more and more application scenarios are incorporating AI to improve existing operations, greatly enhancing the intelligence of work and increasing efficiency. In fields such as equipment repair or insurance claims assessment, it is often necessary to identify the external damage to equipment and then perform repairs and damage assessments based on the identification results.

[0055] By introducing artificial intelligence (AI) technology into the automatic identification of equipment surface damage, and using a pre-built defect identification model to identify surface damage, vehicle owners or damage assessors can upload images of damaged areas to their mobile phones. The system automatically identifies the damaged parts and damage type, enabling rapid identification and assessment, and significantly improving the efficiency of surface damage identification. However, the inventors of this application discovered that when using AI technologies such as defect identification models for defect identification, inaccurate results often occur, leading to incorrect diagnostic reports. In practice, users often cannot easily modify the identification results themselves; they can only report errors to the system administrator, who then improves the model in the background to correct the identification results. This method significantly increases the difficulty of using the defect identification model and causes considerable inconvenience to users.

[0056] In view of this, embodiments of this application propose a method, device, and system for identifying defects in target objects. The method involves a defect identification model performing preliminary identification of image information, then returning defect location information to the target object defect identification device. The target object defect identification device then marks the defect based on the original image information using this location information. Furthermore, the device allows users to correct and supplement the identification results of the defect identification model. The device provides a modification interface to the user, receives trajectory information from the user's secondary marking, and performs secondary marking based on the first modification, forming the final target object defect identification result. This embodiment of the application improves upon the returned results of the defect identification model by recording the initial image information and the modification marking separately, enabling users to easily modify the defect identification results. This avoids the problem in existing technologies where users cannot independently modify the target object defect identification results when the defect identification model's results are inaccurate, greatly improving the accuracy and convenience of defect identification results.

[0057] The target defect identification method, device, and system proposed in this application embodiment can be applied to a variety of application scenarios and can be used to identify defects in various target objects, such as automobiles, motorcycles, drones, and other devices that require appearance identification. In this application embodiment, only automobiles are used as an example for illustration.

[0058] Figure 1 shows a schematic diagram of the target defect identification system proposed in this application. The target defect identification system includes a target defect identification device 10 and a server 20. The target defect identification device 10 is usually a terminal device used by the user. The target defect identification device 10 communicates with the server 20 through a network. The target defect identification device 10 is used to provide an interface and interface for interaction with the user on the user side. The server is used to process the information obtained by the target defect identification device 10.

[0059] The target defect identification device 10 can be a mobile phone, a PAD, or a dedicated car diagnostic tool, etc. The target defect identification device 10 is equipped with a dedicated APP, which communicates with the server 20 through its functions. The target defect identification device 10 includes a camera, a display screen, a processor, a memory, and a communication interface. The APP program is stored in the memory and runs through the processor to utilize the camera, display screen, etc.

[0060] Figure 2 shows a structural diagram of the target defect identification device 10 proposed in an embodiment of this application. As shown in Figure 2, the target defect identification device 10 includes: a camera 101, a display screen 103, a processor 102, a memory 106, a communication interface 104, and a communication bus 108.

[0061] The camera 101, display screen 103, processor 102, memory 106, and communication interface 104 communicate with each other via communication bus 108. The memory 106 stores at least one program 110, which causes the processor 102 to execute the target defect identification method proposed in this application embodiment.

[0062] The camera 101, when invoked by the APP, is used to capture images of the target object to be detected and obtain image information. The capturing can be video or pictures. It can capture a single picture or multiple pictures continuously. Alternatively, after capturing a video, the APP can automatically extract relevant pictures from the video. This embodiment of the application does not impose any limitations.

[0063] The display screen 103 can be a touch screen or an LCD screen, and it has the function of interacting with the user. On the one hand, it can receive the operation command information input by the user when the user operates through the display screen. On the other hand, it can also display the information generated by the target defect identification device 10 itself or the information received by the target defect identification device 10 from the server 20 through the display screen.

[0064] The processor 102 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The one or more processors included in the target defect identification device 10 may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0065] Memory 106 is used to store program 110. Memory 106 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0066] Specifically, program 110 can be called by processor 102 to execute the target defect identification method proposed in the embodiments of this application, so that the target defect identification device 10 can interact with the server 20 to complete the identification operation of the target defect.

[0067] Referring again to Figure 1, the server 20 can be a local server or a cloud server. This server runs a defect identification model, which can be a variety of deep learning models, such as a convolutional neural network (CNN) based model or a YOLOv8 deep learning based model. The CNN-based model can automatically learn hierarchical features in image data, from edges and textures to complex shapes and object parts, exhibiting strong robustness. The YOLOv8-based model uses an innovative backbone network and neck structure, combining the advantages of Transformer and CNN to effectively improve image feature extraction capabilities. It demonstrates excellent performance in target detection, achieving high detection accuracy across various datasets, while maintaining high detection accuracy and real-time detection speed. In this embodiment, a suitable deep learning model can be selected for appearance defect identification according to actual needs; no limitation is imposed in this embodiment.

[0068] Before the target object defect recognition system can operate, the defect recognition model needs to be trained on the server 20. The defect recognition model is trained by acquiring a large training dataset, which may include a large amount of vehicle image information. Various defect areas may be identified in the vehicle images. After the defect recognition model is trained, the target object defect recognition device completes the recognition of target object defects through interaction with the server.

[0069] During the operation of the target object defect recognition system, the user first takes a picture of the target object, such as a vehicle, using the target object defect recognition device 10 to obtain image information of the target object, and then sends the image information to the server 20. The server 20 identifies defects in the image information using a pre-trained defect recognition model and sends the defect recognition result to the target object defect recognition device 10. The defect recognition result includes defect location information. Normally, after the defect recognition model completes the identification of defects in the image information, it outputs an image file with defect area identifiers. However, users cannot modify image files. Therefore, in this embodiment, the defect recognition model is modified so that it only outputs defect location information, so that users can calibrate and supplement the defect recognition results later.

[0070] The target object defect identification device 10 receives the defect location information sent by the server 20 through a communication network, and directly identifies the defects in the original captured image information based on the defect location information, generating a first identification image. This first identification image represents the defect identified by the defect identification model. To facilitate subsequent user modifications, the target object defect identification device records both the original image information and the defect location information. After generating the first identification image, the target object defect identification device displays the preliminary identification results on a display screen.

[0071] After viewing the first identification image, the user can inspect it. If the first identification image contains errors or omissions, the user can operate on the first identification image through the display screen of the target object defect identification device to re-identify the omissions or errors. The target object defect identification device acquires the operation trajectory information corresponding to the operation, identifies the first identification image according to the operation trajectory information, and generates a second identification image. The target object defect is then displayed based on the second identification image. The second identification image is the identification result after the user recalibrates the first identification image, and the target object defect identification device displays the calibrated defect identification result on the display screen.

[0072] Through the above embodiments, the target object defect identification system processes the initial image information of the target object and the defect location information obtained after preliminary identification of the image information by the defect identification model separately, displays the preliminary identification results, and allows users to calibrate and modify the defect identification results after preliminary identification to obtain a calibrated and modified second identification image. The second identification image is then used as the final defect identification result. This approach avoids the problem in the prior art where users cannot independently modify the target object defect identification results when the defect identification model's identification results are inaccurate, greatly improving the accuracy and convenience of defect identification results.

[0073] When a target object is identified using the target object defect identification system proposed in this application embodiment, the target object defect identification device runs the target object defect identification method proposed in this application embodiment. Specifically, Figure 3 shows a flowchart of the target object defect identification method proposed in this application embodiment, which includes the following steps:

[0074] Step S110: Acquire image information of the target object;

[0075] This target defect identification device has multiple functions, including vehicle exterior damage inspection, tire inspection, vehicle interior and exterior inspection, engine compartment inspection, battery performance inspection, and vehicle undercarriage inspection. Users can start the target defect identification method process by logging into the APP on the target defect identification device and opening the vehicle exterior damage detection function.

[0076] After activating the vehicle exterior damage detection function, the first step is to capture an image of the target object using a camera. Multiple images can be taken, and the most suitable one can be selected for upload.

[0077] Step S120: Send the image information to a pre-trained defect recognition model so that the defect recognition model can identify defects in the image information;

[0078] After acquiring the image information, the captured photos can be compressed and sent to the server via the network. The defect recognition model on the server will then identify the defects in the image information.

[0079] Step S130: Receive the defect identification result generated by the defect identification model, wherein the defect identification result includes defect location information;

[0080] In the results of the defect identification model, each image may have multiple damage points, and each damage point has a returned array of coordinate values. The array of coordinate values ​​is used to describe the area range of the damage point and represents the location information of the defect.

[0081] Preferably, in order to reduce the amount of data transmitted between the server and the target defect identification device and improve data transmission efficiency, the coordinate value array may only include the coordinate values ​​of the edge points of the area where the damage point is located. In this way, only the coordinates of the edge points need to be connected to form a line to represent the location of the damage point. It is not necessary to transmit all the coordinate values ​​of the area where the damage point is located, which greatly reduces the amount of data transmitted between the target identification device and the server and improves the identification efficiency.

[0082] Meanwhile, in order to further reduce the amount of data transmitted between the target object recognition device and the server, the coordinate values ​​of the edge points of the area where the damage point is located can be sampled. For example, a coordinate value can be extracted every 30 pixels, which can further reduce the amount of data on the defect location information.

[0083] Step S140: Identify the defects in the image information according to the defect location information to generate a first identified image;

[0084] Because the defect recognition model feeds back coordinate values ​​instead of returning directly labeled images, the target object defect recognition device can identify defects in the image information based on the defect location information, and store the image information and defect location information separately, allowing users to modify and calibrate the recognition results of the defect recognition model.

[0085] Step S150: In response to the operation on the first identifier image, obtain the operation trajectory information corresponding to the operation;

[0086] If the user is not satisfied with the recognition results generated by the defect recognition model, or if the recognition results generated by the defect recognition model are incomplete, the user can modify the first identification image through the display screen. When the display screen is a touch screen, the user can directly draw lines around the defect area with their finger to re-mark the defect area.

[0087] Step S160: Mark the first identification image according to the operation trajectory information to generate a second identification image;

[0088] After the user re-marks the defective area, the target object defect recognition device will display the drawn area on the screen. Simultaneously, to reduce the data processing load and improve the processing efficiency of the target object defect recognition device, when recording the operation trajectory, the device can obtain multiple coordinate points corresponding to the operation trajectory information based on a preset pixel spacing threshold; and then mark the first marking image based on these multiple coordinate points to generate a second marking image. For example, if the preset pixel spacing threshold is 30 pixels, the target object defect recognition device samples the coordinate value of one coordinate point every 30 pixels and generates the operation trajectory based on the sampled coordinate values.

[0089] It should be noted that users can modify or add multiple operation trajectories multiple times. The target defect identification device will acquire all operation trajectory information and merge it with the first identification image to form the second identification image.

[0090] Step S170: Display the defects of the target object according to the second identification image.

[0091] The second identification image represents the target defect identification result calibrated by the user, and the target defect identification device displays the final target defect identification result on the display screen.

[0092] Through the above embodiments, the target object defect identification device processes the image information of the target object and the defect location information obtained after preliminary identification by the defect identification model by executing the target object defect identification method, displays the preliminary identification results, and allows users to calibrate and modify the defect identification results after preliminary identification to obtain a calibrated and modified second identification image. The second identification image is used as the final defect identification result. This method avoids the problem in the prior art that users cannot independently modify the target object defect identification results when the defect identification model identification results are inaccurate, and greatly improves the accuracy and convenience of defect identification results.

[0093] Furthermore, in some embodiments, considering the characteristics of the defect recognition model, it is necessary to continuously train to improve the accuracy of model recognition. Therefore, in this embodiment, after the user completes the calibration of the defect recognition model's recognition result and generates a second identifier image, the second identifier image can be sent to the server so that the defect recognition model can perform self-calibration training based on the calibrated second identifier image, so as to further improve the recognition accuracy of the defect recognition model.

[0094] To more clearly illustrate the process of the target defect identification system and target defect identification device operating the target defect identification method proposed in this application embodiment, please refer to Figure 4, which shows the signaling flowchart in the target defect identification system.

[0095] In practical applications, the amount of data in the images requiring appearance recognition is very large. In this embodiment, the server is divided into an AI server 21 and a file server 22, as shown in Figure 1. The AI ​​server 21 is mainly used to run and train the defect recognition model and interact with the target object defect recognition device 10. The file server 22 is mainly used to store image information and interact with both the AI ​​server 21 and the target object defect recognition device 10. In this embodiment, the target object defect recognition device 10, AI server 21, and file server 22 are respectively configured to execute the above-mentioned target object defect recognition method, specifically including:

[0096] Step S201: The target object defect identification device takes a picture using a camera;

[0097] Figure 5 shows the shooting interface of the target defect recognition device. This shooting function is used to acquire the original material images for AI recognition and supports functions such as focus, flashlight, and photo preview. As shown in Figure 5, after each image is taken, the captured image will be displayed at the bottom of the shooting interface. If you are not satisfied with the shooting effect, you can delete the image by clicking the "x" button.

[0098] Step S202: The target object defect identification device compresses the acquired image information;

[0099] After taking a picture of the target object, the target object defect identification device needs to send the captured image to a file server. In order to improve the image transmission efficiency, in this embodiment of the application, the image can be packaged and compressed into a .zip or .rar file.

[0100] Step S203: The target object defect identification device sends a pre-signed address retrieval message to the AI ​​server;

[0101] In this embodiment of the application, in order to improve the working efficiency of the AI ​​server and reduce the burden on the AI ​​server, the AI ​​server and the file server are divided into two separate servers. When the target object defect identification device uploads image information, it needs to upload the image to the file server. Therefore, before uploading the image to the file server, it needs to send a pre-signature address acquisition message to the AI ​​server to obtain the address of the file server.

[0102] Step S204: The AI ​​server returns the pre-signed address information;

[0103] In response to the pre-signature address acquisition message from the target defect identification device, the AI ​​server returns pre-signature address information to the target defect identification device after authentication and authorization. The pre-signature address information includes the address information of the file server.

[0104] Step S205: The target object defect identification device uploads image information to the file server;

[0105] The target defect identification device uploads the compressed image information to the file server so that the AI ​​server can access it to analyze the image information.

[0106] Step S206: The file server sends a message indicating successful upload;

[0107] Once the file server successfully receives the file sent by the target defect identification device, it returns an upload success message to the target defect identification device.

[0108] Step S207: The target object defect identification device sends an AI analysis request to the AI ​​server;

[0109] The AI ​​parsing request can be initiated by the user through the display screen of the target object defect identification device, sending the AI ​​parsing request to the AI ​​server under the user's instruction; or the target object defect identification device can automatically send the AI ​​parsing request to the AI ​​server after receiving the upload success message from the file server.

[0110] Step S208: The AI ​​server obtains image information from the file server.

[0111] Step S209: The file server returns image information to the AI ​​server.

[0112] Step S210: The AI ​​server calls the AI ​​engine to identify defects;

[0113] AI servers can analyze and predict image information using YOLOv8 models or convolutional neural network models to identify defects.

[0114] Step S211: The AI ​​server returns the defect location information;

[0115] The defect location information is an array of coordinate values ​​corresponding to the defect area. This array of coordinate values ​​describes the area range of the damaged point, thereby indicating the defect location information.

[0116] Step S212: The user manually modifies the AI ​​recognition result using the target object defect recognition device;

[0117] After the target object defect identification device receives the defect location information sent by the AI ​​server, it marks the defect area on the original image information obtained by the camera according to the defect location information, forming a first marked image, which is used to display the identification result of the AI ​​server to the user.

[0118] As shown in Figure 6, the recognition result preview window 122 in the lower area of ​​the main display window 121 displays multiple recognition results from the AI ​​server. When the user selects an image in the recognition result preview window 122, it will be displayed in the main display window 121, showing the defect area formed based on the defect location information generated by the AI ​​server. When the user selects to delete a recognition result generated by the AI ​​server, they can click the inverse selection button on the recognition result preview window 122 to delete the corresponding first-identification image, that is, delete the identification generated based on the defect location information returned by the AI ​​server. Only the original image information is displayed in the main display window.

[0119] Furthermore, referring to Figure 6, to the right of the main display window 121, there is also an attribute window 123. This attribute window is used to identify the type of defect in the defect identification result. The attribute window includes defect type information, which indicates the type of defect, such as scratches, dents, cracks, wrinkles, or missing parts. Furthermore, the defect identification result also includes component type information, which identifies the location of the defect, such as the front, left, right, rear, upper, and wheel.

[0120] After the AI ​​server parses the image information, the returned defect identification results will automatically generate the defect type information and the component type information based on the shape and other information of the defect. When the AI ​​server's identification results are displayed through the main display window 121, the defect type information and the component type information will be displayed in the attribute window 123.

[0121] When a user needs to modify the recognition result of an AI server, they can click the "Manual Drawing" button in the operation window 124 on the right. This will generate a new label, which the user can directly mark in the main display window 121 by drawing lines. At the same time, attribute information, including defect type information and component type information, will be added to the newly added manual drawing operation. The user can select the corresponding defect type information and component type information through the drop-down menu.

[0122] After modifying the first identifier image, the modification result can be saved.

[0123] Step S213: The user confirms the modification result through the operation interface of the target object defect identification device;

[0124] After the user has completed the modification and confirmation of all the first identifier images, they can select the final second identifier image in the recognition result preview window 122. If they need to delete or not select certain images, they can invert the selection to abandon the use of that identifier image.

[0125] Once the user has selected and finalized the second identification image, they can click the "OK" button to complete the final calibration of the defect identification image, which will then be displayed in the interface shown in Figure 7.

[0126] In Figure 7, defects in the second identification image can be displayed and statistically analyzed according to their attribute information. For example, they can be statistically analyzed according to defect type information. When a user selects to perform a query operation based on defect type information, the target object defect identification device traverses the defect identification results and the operation trajectory information to generate query results. As shown in Figure 7, the number of defects is displayed according to defect type information.

[0127] Step S214: The target object defect identification device sends the modification results back to the AI ​​server;

[0128] After the target object defect identification device completes the identification of the target object defect, it will save the second identification image locally to generate an identification report and trace related historical defect information. Simultaneously, to improve the accuracy of the defect identification model, the second identification image can also be sent to the AI ​​server so that the defect identification model can perform self-calibration training based on the second identification image.

[0129] Step S215: Complete the recognition and display the recognition result on the display screen;

[0130] Figure 7 shows the final defect identification result.

[0131] In summary, the target defect identification system, device, and method provided in this application improve upon the return results of the defect identification model by recording the initial image information and modification identifiers separately. This facilitates the modification of defect identification results and avoids the problem in existing technologies where users cannot independently modify the target defect identification results when the defect identification model's results are inaccurate. This significantly improves the accuracy and convenience of defect identification results. Furthermore, sending a second identifier image, which has been manually identified and calibrated, to the defect identification model allows for retraining and self-calibration, further enhancing the model's accuracy.

[0132] This application also provides a computer-readable storage medium storing executable instructions. When the executable instructions are run on the target defect identification device, the target defect identification device performs the operation of the target defect identification method provided in any of the above embodiments.

[0133] This application also provides a target defect identification program, which is used to execute the target defect identification method provided in the above embodiments.

[0134] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of this application are not directed to any particular programming language. It should be understood that the content of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of this application.

[0135] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0136] Similarly, it should be understood that, in order to simplify this application and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of this application, various features of the embodiments of this application are sometimes grouped together into a single embodiment, figure, or description thereof.

[0137] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying abstract and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying abstract and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0138] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying defects in a target object, characterized in that, include: Acquire image information of the target object; The image information is sent to a pre-trained defect recognition model so that the defect recognition model can identify defects in the image information; Receive the defect identification result generated by the defect identification model, wherein the defect identification result includes defect location information; Based on the defect location information, the defects in the image information are identified to generate a first identified image; In response to an operation on the first identified image, obtain the operation trajectory information corresponding to the operation; The first identification image is identified based on the operation trajectory information to generate a second identification image; The defects of the target object are displayed based on the second identification image.

2. The method according to claim 1, characterized in that, The step of identifying the first identification image based on the operation trajectory information and generating a second identification image includes: Multiple coordinate points corresponding to the operation trajectory information are obtained according to a preset pixel spacing threshold; The first identification image is identified based on multiple coordinate points to generate a second identification image.

3. The method according to claim 1, characterized in that, The step of identifying defects in the image information based on the defect location information and generating a first identified image includes: In response to the inverted selection operation on the first identifier image, the identifier on the first identifier image is deleted.

4. The method according to any one of claims 1-3, characterized in that, The defect identification result also includes defect type information, which is used to indicate the defect type of the defect area corresponding to the defect location information. The step of obtaining operation trajectory information corresponding to the operation in response to an operation on the first identifier image further includes: Obtain the defect type information corresponding to the operation trajectory information; The method further includes: In response to a query operation targeting the defect type information, the defect identification results and the operation trajectory information are traversed to generate query results.

5. The method according to claim 4, characterized in that, The method further includes: The second identifier image is sent to the defect recognition model so that the defect recognition model can perform self-calibration training based on the second identifier image.

6. A target object defect identification system, characterized in that, The system includes a target defect identification device and a server; The target defect identification device acquires image information of the target object and sends the image information to the server; The server runs a pre-trained defect recognition model, which is used to identify defects in the image information and send the defect recognition results to the target object defect recognition device. The defect recognition results include defect location information. The target object defect identification device receives the defect location information and identifies the defects in the image information according to the defect location information, generating a first identification image; The target object defect identification device is further configured to, in response to an operation on the first identification image, acquire operation trajectory information corresponding to the operation; identify the first identification image according to the operation trajectory information to generate a second identification image; and display the target object defect according to the second identification image.

7. The system according to claim 6, characterized in that, The server includes an AI server and a file server; The target defect identification device is communicatively connected to the AI ​​server and the file server, respectively. The target defect identification device is also used to send a pre-signed address acquisition message to the AI ​​server; The AI ​​server sends pre-signature address information to the target defect identification device based on the pre-signature address, and the pre-signature address information includes the file server address. The target defect identification device sends the image information to the file server according to the file server address.

8. The system according to claim 7, characterized in that, The target defect identification device is also used to send an AI analysis request to the AI ​​server in response to the AI ​​analysis operation; The AI ​​server obtains the image information from the file server according to the AI ​​parsing request in order to identify defects in the image information.

9. The system according to claim 8, characterized in that, The target defect identification device is also used to save the second identification image locally and send the second identification image to the AI ​​server so that the defect identification model can perform self-calibration training based on the second identification image.

10. A target object defect identification device, characterized in that, include: The system includes a camera, a display screen, a processor, a memory, a communication interface, and a communication bus. The camera, display screen, processor, memory, and communication interface communicate with each other via the communication bus. The camera is used to capture images of the target object and obtain its image information; The display screen is used to provide an interactive interface with the outside world and to display the defect identification results; The memory is used to store at least one program that causes the processor to perform the operation of the target defect identification method as described in any one of claims 1-5.

11. A computer-readable storage medium, characterized in that, The storage medium stores at least one program, which, when run on the target defect identification device, causes the target defect identification device to perform the operation of the target defect identification method as described in any one of claims 1-5.

Images